I had a very interesting conversation with a good friend a couple of weeks ago about causation and inferring causal relationships haphazardly.
One of the questions that stuck in my head was this:
Are big companies big because they advertise a lot?
Or do they advertise a lot because they are big?
Sounds like a chicken or the egg situation: which came first, the chicken or the egg?
(My friend thinks it's neither - it's the archaeopteryx. To which I argued, "That was a bird." To which she said, "Nope, that's the ancestors of birds." ... I have weird friends, I know.)
But it is something worth pondering - because at the root of this question is the question of causality.
Causality is supposed to be simple:
If A happens, B happens.
Without A, B does not happen.
Therefore, A causes B.
Or is it?
What if there is a lurking variable C that is somehow related to both A and B, such that:
If A happens, C happens.
If C happens, B happens.
If A does not happen, C does not happen.
If C does not happen, then B does not happen.
Can we still consider arrive the same conclusion that "A causes B"?
Let's make it even more complicated:
If A and C happen, B happens.
If A happens, B does not happen.
If C happens, B does not happen.
If A and C don't happen, B does not happen.
Therefore A and C happening cause B happening.
But what if A and C are themselves related somehow? Worse, what if A causes C? Or what if there was another variable D that we have not considered that is making the presence-absence of A and of C? Can we still arrive at the same conclusion?
I am no logician. And I am pretty sure that there is an answer to these.
I can already hear my psych professor going "Do a proper experiment in a laboratory setting eliminating all potential confounding variables, and measure properly - and do it several times - and get others to do it several times too". But sometimes, experiments are not easy to do.
Even the most stringent of experiments still tend to have some flaw in it.
So what now?
All I can say, for now, is -